<_whitenotifier-7>
[scopehal-apps] azonenberg a39c2aa - Fix potential hang during session destruction
<azonenberg>
We can now trigger and acquire data from a scope. It doesn't go anywhere as there's no way to create filters, there's no history, and there's no rendering
<azonenberg>
but it does stuff. Data from the picoscope piles up even when we're popping waveforms at 60 FPS
<azonenberg>
i'm going to add some metrics and stats collection shortly so we can see exactly how fast we can pull data
<azonenberg>
But most important is that profiling shows the main thread barely doing anything. which is exactly what we want, to keep things nice and responsive
<bvernoux>
azonenberg, the famous m_nickname issue when it is empty is everywhere ;)
<bvernoux>
more I fix more I find new ones
<bvernoux>
I'm not even sure it is always invalid to have an empty m_nickname ...
<azonenberg>
empty nickname should never be allowed
<azonenberg>
the fix is to catch it when you connect to the scope
<azonenberg>
the connection dialog should reject it
<bvernoux>
yes
<d1b2>
<ehntoo> Looks like Rigol is hopping on the 12-bit converter train. Wonder what the internal architecture is like. https://www.rigolna.com/news/2022/8764/
<d1b2>
<ehntoo> ENOB is only specified at ">8", so...
<bvernoux>
yes Rigol ENOB is very low on 8bits down to 5bits
<bvernoux>
the lowest on market IIRC
<bvernoux>
even down to 4 ENOB for MOS5K at 8GSPS
<d1b2>
<ehntoo> Ooh, it has a USB 3 device port.
<bvernoux>
yes like Gigabit Ethernet on MSO5K where you cannot push more than 5MBytes/s of waveform ;)
<bvernoux>
it is clearly their main issue the SCPI waveform data which is dead slow
<bvernoux>
a good things on their new MDO4000 is they have 50Ohm or 1GOhms input selectable
<d1b2>
<ehntoo> The SCPI stack on the MSO5K is definitely not optimized for speed. I'm working on that.
<bvernoux>
with 10MHz Ref Clock
<bvernoux>
MSO5K need a new firmware
<bvernoux>
you cannot optimize anything more I tried everything
<bvernoux>
lot of latency are in their architecture
<d1b2>
<ehntoo> I know. 🙂
<bvernoux>
MSO5K have clearly the capability to push 20WFM/s even at 100kpts
<bvernoux>
but Rigol do not want to spend time on that
<bvernoux>
I was in contact with their R&D in Germany and they promised stuff each 6months
<bvernoux>
and at end after more than 1 year they said the main R&D have other priorities
<bvernoux>
They even told they have never specified anything related to SCPI waveform data speed
<bvernoux>
the question is why to put a Gigabit Ethernet in that case
<bvernoux>
as they do not even reach the speed of 100Mbps Ethernet ...
<bvernoux>
I'm pretty sure they could have done a quick fix for fast SCPI
<bvernoux>
by just locking the UI to have full access to SCPI to retrieve waveform
<bvernoux>
as it is clearly full of issue between UI & SCPI
<bvernoux>
I suspect some ugly mutexes linked between UI & SCPI to explain why it is so slow to retrieve data
<bvernoux>
some buggy mutexes doing timeout ;)
<bvernoux>
Anyway their Rigol MSO5K are just the cheapest most advanced scope(when fully unlocked) available for <1KEuros
<bvernoux>
Even if the noise floor is quite high we do not use oscilloscope to measure voltage level anyway ;)
<bvernoux>
So that use case is not for those cheap scope
<bvernoux>
Will be interesting to ask to Chris Armstrong (Director of Rigol USA) what is the speed to retrieve waveform over SCPI with their brand new MDO4000 ;)
<bvernoux>
Does they exceed 1WFM/s ;)
<bvernoux>
their UI seems different than on MSO5K
<bvernoux>
they really need something like ngscopeclient ;)
<bvernoux>
they are using Cortex-A72, 1.8 GHz, hexa-core
<bvernoux>
with Android
<bvernoux>
it is first time I see an Oscilloscope using Android
<_whitenotifier-7>
[scopehal-apps] bvernoux edited pull request #500: Fix crash, use a default nickname for each instrument (if the nickname is empty force user to set a nickname) - https://github.com/glscopeclient/scopehal-apps/pull/500
<_whitenotifier-7>
[scopehal-apps] bvernoux edited pull request #500: ngscopeclient Fix crash, use a default nickname for each instrument (if the nickname is empty force user to set a nickname) - https://github.com/glscopeclient/scopehal-apps/pull/500
<d1b2>
<karlp> so, I'm reading the latest pdf installation notes, and have latest source, and just running cmake at the start fails: https://paste.jvnv.net/view/lch3q lots of warnings about "missing cmakelists.txt" and it doesn't seem skippable?
<d1b2>
<karlp> is there something obvious I'm missing?
<miek>
sounds like the submodules aren't checked out
<d1b2>
<karlp> I've pulled all the submodules, so it's (hopeuflly) not anything that simple
<miek>
did you pull them recursively?
<d1b2>
<karlp> hrm, I did a git submoduel update --init.
<d1b2>
<karlp> let me double check if that went deep enough
<miek>
add a --recursive to that :)
<d1b2>
<karlp> ok, that does look better now 🙂
<d1b2>
<karlp> lets see if this gets further...
<d1b2>
<karlp> annnnd it aborts 🙂
<d1b2>
<karlp> lol, it crashed gdb too 🙂
<d1b2>
<karlp> ok, this obviously needs a different night then.
<azonenberg>
karlp: what platform are you testing on?
<d1b2>
<karlp> fedora36. I'll look again tomorrow, spent enough time on tech stuff today 🙂
Guest98 has joined #scopehal
<Guest98>
Hi. Nice to meet you.
<Guest98>
I'm a beginner who uses an oscilloscope as a hobby.
<Guest98>
Looking for a way to connect to tds3000b (lan) in glscopeclient.
<Guest98>
Cannot connect to tektronix in the menu.
<azonenberg>
Guest98: So far the only Tek scopes our driver supports is the modern MSO5/6 series, and possibly MSO2/4 which use a similar command set but AFAIK nobody has tested
<azonenberg>
i have no idea if the TDS3000 are at all similar or not
<azonenberg>
The programming guide is easy to google up, so if someone wanted to write a driver and had access to a scope it probably wouldn't be too hard to get basic functionality working